Partition exchange takes an X lock out on both tables and invalidates cursors on both of them. That does not change with the # of [sub]partitions. However, there is row cache overhead (for obj$) that does increase linearly. On Fri, Feb 11, 2011 at 12:57 PM, <paul.baumgartel@xxxxxxx> wrote: > Oracle11g 11.2.0.2. on Exadata. > > Many waits for library cache locks are observed during a run in which many > processes exchange subpartitions in one particular table. > > The application ensures that two sessions will not try to exchange the same > subpartition at once. We understand that the waits for library cache locks > are normal and expected. > > The table in question contains 90,000+ subpartitions. The question has come > up of whether the time that Oracle spends acquiring and releasing the > library cache locks is in any way proportional to the number of > subpartitions. > > Thoughts? -- Regards, Greg Rahn http://structureddata.org -- //www.freelists.org/webpage/oracle-l